Two Rows Became One Row

On 9 June 2025 Pew Research Center published a world religion table with a single line for other religions, putting it at roughly 157 million people.

The earlier report had not used that line. Its 2010-data table listed some 400 million folk religionists and 58 million adherents of other religions, as separate rows. Add the two old rows together and the population appears to have fallen from about 458 million to 157 million in a decade.

Nothing of the sort was established.

The 2025 report drew the bucket differently.

The collapse is an event in the reporting structure before it is an event in the world.

That is the tame version of an error made routinely after a schism. A spreadsheet wants one predecessor, one successor, and a second line beginning at the separation date. A split can instead leave two bodies each carrying parts of the old institution, and the line that looks newly born may be holding inherited members, offices, assets, mandates and identifiers.

Calling it a new series does not discover anything. It selects an allocation rule, and then hides the selection.

A row is not the population it counts #

Pew’s figures are stitched from thousands of censuses, surveys and population registers across two hundred-odd countries, harmonised to a common date. Neither round is a reading taken on a single day, and the national instruments behind them do not ask the same question as each other.

The Chinese change makes that concrete. For the 2020-data round Pew adopted a zongjiao religious-identity measure in place of earlier proxies, and more than sixty national censuses were delayed by COVID-19 besides.

A line can move because the respondent moved, or the fieldwork date, or the question, or the bucket. A single plotted series conceals every one of those unless the definitions travel alongside it.

The folk-religion row is the cleanest case, because there is no dispute at all among the people being classified. The 2010 report separated folk religionists from other religions. The 2025 report consolidated a structurally similar space into one row. No subtraction between the two tables can say how much practice disappeared, because the predecessor population was assigned to different successors before anyone started comparing.

A corporate spin-off is the same bookkeeping problem in a harder setting. A vendor that starts a fresh history for one company and hands the entire old customer base to the other has made a continuity decision. It may well be the right one. It is not neutral, and it cannot be audited afterwards if the transfer rule was never written down.

Two ledgers, kept by different people #

Ahmadiyya is the sharper warning, because there the boundary changes the count.

Ahmadis identify as Muslim.

Pakistan’s 1974 constitutional amendment excludes Ahmadis from the constitutional definition of Muslim, and the Organisation of Islamic Cooperation had declared the year before that Ahmadiyya was not linked to Islam.

Pakistan’s official census counts a few hundred thousand self-reported Ahmadis. Independent estimates commonly run to several million.

That gap is not measurement error around a shared object. It records two classification systems operating under very unequal consequences, and a census line can be entirely precise about what its form permits a person to say while being useless as a count of affiliation.

The Latter-day Saint case separates the same records without any legal enforcement involved. A 2016 Pew survey found that almost all Latter-day Saints considered themselves Christian, and about half the American public agreed. Several major churches do not recognise the LDS Church as part of the historic Christian tradition.

There is no single figure called Christian until the institution, the respondent and the researcher have agreed whose claim takes priority.

So an inherited label cannot be read as a settled lineage. It can mean a self-description, a legal status, a doctrinal recognition, a census response, or a claim to an office. Those are not competing spellings of one field. They are separate fields, and they can agree for decades and then part at precisely the moment the history becomes expensive.

Keep the fork the evidence left you #

Biblical criticism has had a discipline for this since the eighteenth century, and its habit is the transferable part.

Jean Astruc treated the alternating divine names and the doublets in Genesis as evidence of earlier documents rather than as a puzzle to be smoothed away. The flood account gives both forty days and a hundred and fifty, and both one pair of every animal and seven pairs of the clean ones. The documentary hypothesis does not resolve those into a cleaner narrative. It keeps them as evidence about transmission.

The later revision matters just as much. Three works published in the 1970s argued that the Jahwist material belonged to the Babylonian exile rather than to Solomon’s court, moving a source by four centuries.

The prior sequence had looked entirely settled.

Its inputs had not.

That is the restraint worth taking. Record the split, the claimants and the transfers before calculating a continuous series. Keep a fork wherever the evidence leaves one standing. Where the record does not choose an allocation rule, run the result under more than one and report both.

Two Pew rows became one Pew row, and three hundred million people appeared to leave a religion that none of them had left.

A schism does the same thing to an institution’s history, with one difference: the inherited members, offices and obligations are real, and somebody is still carrying them.

The fork is much harder to draw than the line. It is also the only place the exposure is still recorded.